Deep excavation and station boxes

A metro station box is a hole 20–30 m deep and 200 m long, dug between buildings that must not move, in ground that is usually full of water, under a road that has to stay open.

Almost nothing about it is about strength. It is about movement and water.

A deep urban excavation for a metro station box seen from street level, diaphragm walls on all four sides with several levels of tubular steel props and waler beams spanning across the opening, an excavator working on the partly excavated floor far below, and city buildings and a live road crowding the edge. — Four levels of props and the road still open a few metres from the wall. Nothing in this picture is governed by strength — it is governed by how much the wall is allowed to move before the buildings on the left notice.

The wall

The wall does three jobs, and which type you choose depends on how many of them it has to do.

Wall type	Retains	Cuts off water	Becomes permanent	Typically
Sheet piles	Yes	Fairly	Rarely	Shallow, soft ground, recoverable
Contiguous bored piles	Yes	No — gaps between piles	Sometimes	Above the water table, cohesive ground
Secant piles	Yes	Yes — piles intersect	Often	Medium depth, tight sites
Diaphragm wall	Yes	Yes	Yes	Deep station boxes, 20 m+

Four excavation support wall types in plan and section. Sheet piles shown as an interlocking corrugated line. Contiguous bored piles shown as a row of circles with visible gaps between them and water seeping through. Secant piles shown as overlapping circles with alternating soft primary and reinforced secondary piles forming a continuous line. Diaphragm wall shown as rectangular panels with a stop end between them and a reinforcement cage, with a section showing bentonite support fluid in the trench and the fluid level above the external groundwater level. — The wall choice is a water question before it is a strength question. Contiguous piles leave gaps on purpose; secant piles and diaphragm walls do not.

The trench is held up by a liquid

A diaphragm wall panel is excavated as an open trench full of bentonite or polymer support fluid. The fluid is the only thing holding the trench walls apart.

It works because the fluid level is kept above the external groundwater level, so there is a net outward pressure and a low-permeability cake forms on the trench face.

Lose that head — a burst supply line, a leak into a permeable seam, a stoppage that lets the level drop — and the panel collapses. Not slowly. Fluid level is monitored continuously on every panel, and there is no such thing as a brief interruption.

Secant piles solve the same problem differently: soft or hard primary piles are bored first, then secondary reinforced piles are bored through them, cutting into both neighbours. The overlap is the water cut-off, and verticality tolerance is what guarantees the overlap still exists at the toe.

Propping: bottom-up against top-down

The wall spans between supports. With excavation levels descending, those supports arrive one at a time.

Bottom-up. Excavate to the first prop level, install and preload props, excavate to the next, and so on. Cast the base slab, then build up, transferring the wall support onto the permanent slabs and removing the temporary props as you go.

Simple, familiar, well understood.
The box is full of steel props, so plant access and concrete placement are awkward.
The surface is occupied for the whole duration.

Top-down. Install the walls and the plunge columns. Cast the roof slab on the ground, reinstate the road over it, then dig out beneath it through openings, casting each permanent slab as you reach its level. Every permanent slab is a prop.

No temporary props in the box.
The road comes back early — decisive on a city corridor.
The superstructure can rise while the basement is still being excavated.
Harder: muck out through holes, low headroom, and the plunge column verticality and the slab-to-column joints are unforgiving.

Ground anchors are the third option — drilled back into the soil behind the wall and stressed. They leave the box completely clear, but they need the ground behind the wall to be available, which under an adjacent building or an existing tunnel it usually is not, and they need wayleaves from whoever owns it.

Water, from below

Two base failure mechanisms compared in section. Base heave shows a deep excavation in soft clay with the weight of retained ground outside pushing soil around the wall toe and up into the excavation, with the failure surface drawn as a curve passing under the toe. Hydraulic base failure shows an excavation whose base sits above a confined sand aquifer, with the uplift water pressure acting on the underside of a clay plug and the weight of that plug resisting it, and the block lifting bodily when pressure exceeds weight. Each is annotated with its cure: deeper toe or base treatment for heave, relief wells or depressurisation for uplift. — Two different failures that both lift the base. The first is a soil strength problem, the second is a simple weight comparison — and the cures have nothing in common.

Base heave. Dig out the inside while the full weight of ground remains outside, and the difference is a bearing capacity problem running backwards. In soft clay the soil flows around the wall toe and the base rises. The check is a stability number based on the undrained strength and the excavation depth; the fixes are driving the toe into competent material, ground improvement below formation, or a thicker base slab cast earlier.

Hydraulic base failure. If a confined aquifer sits below the excavation, its pressure acts on the underside of the soil plug above it. Compare weight against uplift:

Factor of safety  =  (γ_soil · d)  /  (γ_water · h_w)

d   = thickness of the soil plug between formation and the aquifer
h_w = piezometric head in the aquifer above the top of that layer

When that ratio approaches 1, the plug lifts as a block. No amount of propping helps — the cures are relief wells, depressurising the aquifer, or extending the cut-off wall through it.

Instrumentation

Every deep excavation is designed with a predicted movement. Monitoring exists to check the real one against it.

Instrument	Measures
Inclinometers in the wall	Deflection profile with depth
Prop load cells or strain gauges	Actual prop load against design
Ground settlement arrays	The trough behind the wall
Piezometers	Pore pressure inside and outside the cut-off
Building monitoring points, crack gauges	What the neighbours are doing
Automated total stations	Continuous, unattended, out of hours

Where this fits

The station box and the running tunnel are the same corridor and the same ground — see tunnelling and segmental lining for the drive between them, piling and deep foundations for the wall and plunge column installation, and soil investigation and compaction for the data all of this rests on.

What to take away

The wall type is chosen on water cut-off first; contiguous piles deliberately leave gaps.
A diaphragm wall trench is held open by fluid whose level must stay above the groundwater — continuously.
Top-down uses the permanent slabs as props and gives the road back early; bottom-up is simpler but fills the box with steel.
Serviceability governs. Install props early, preload them, never over-excavate below a prop level.
Prop removal sequence is designed, not decided on site.
Base heave is a soil strength failure; hydraulic base failure is a weight-versus-uplift sum. Different cures.
Dewatering settles ground well outside the site, and timber-piled buildings are the worst case.
Trigger levels need pre-agreed responses, and readings worse than predicted mean the prediction was wrong.